Personal Product AI-Powered

Gooey Finance

A personal-finance scheduling app for planning recurring income and expenses across shared households, with AI-driven transaction categorization and analysis — shipped as a web app and a full-parity Flutter mobile client.

Highlights

  • AI-driven transaction categorization and spending analysis
  • Cross-platform: a React web app and a Flutter mobile client at full feature parity, sharing one cookie-based auth model
  • A recurrence engine for scheduled debits/credits with flexible rules, per-occurrence skipping, and weekend handling
  • Invite-only auth across password, OAuth, and magic-link flows, with per-user runtime feature flags and a filterable admin audit log

Skills

AI/LLM integrationFull-stack + mobile deliveryBackend & data-model architectureAWS / infrastructure-as-codeAuthentication & identityTest-driven engineering

Overview#

Gooey Finance is a personal-finance scheduling app I founded and build. It helps individuals and households plan cash flow by scheduling recurring debits and credits against their accounts, rather than just recording what already happened. It also uses AI to categorize transactions and analyze spending, so the numbers turn into insight without manual tagging. It ships as a Fastify + PostgreSQL API, a React 19 + Vite web app, and a Flutter mobile client at full feature parity — all deployed on AWS via Terraform Cloud.

The Problem#

Most budgeting tools track the past; Gooey Finance plans the future. Modeling scheduled and recurring transactions so a household can see where its balances are heading requires a real recurrence engine and a shared-ownership model — not just a flat transaction list.

My Role#

Founder and senior software engineer — product, backend, web, mobile, and infrastructure.

Architecture & Approach#

  • API. Fastify + TypeScript on PostgreSQL, with a domain spanning users, households, finance accounts, schedule items, and recurring rules. Routes are concern-grouped, and there is no public registration — users enter only by admin invite.
  • Web. A React 19 + Vite single-page app.
  • Mobile. A Flutter client (Android + iOS) at full feature parity with the web app — scheduler, transactions, recurrence builder, households, notifications — reusing the same cookie-based auth through a persistent cookie jar rather than bearer tokens.
  • Infrastructure. Terraform Cloud–managed AWS (ECS Fargate, RDS Postgres, S3 + CloudFront), deployed through GitHub Actions using OIDC federation.

Technical Highlights#

  • AI-driven categorization & analysis. Transactions are automatically categorized and analyzed by an AI layer, turning raw debits and credits into labeled, summarized spending insight without the user hand-tagging every line.
  • Recurrence engine. Recurring rules generate schedule items from an interval and unit, days-of-week, end rules, weekend handling, and skip dates, with per-occurrence skipping — the core scheduling logic of the product.
  • Multi-channel invite-only auth. Email/password, Google, Facebook, and magic-link sign-in all converge on a single account, with OAuth identities linked by verified email.
  • Per-user runtime feature flags enforced server-side — gated routes return data only when the caller has the flag, so features aren't merely hidden in the UI — with an admin toggle menu.
  • Shared households with email- or link-token invitations.
  • Engineering discipline. A large automated test suite (800+ tests across API, web, and mobile) and strict PostgreSQL-version pinning kept aligned across local, CI, and production.

Skills Demonstrated#

AI/LLM integration for transaction categorization and spending analysis, full-stack and cross-platform mobile delivery, backend and data-model architecture, AWS infrastructure-as-code, multi-provider authentication and identity, and disciplined test-driven engineering.